Six months ago, Moonshot AI was a wounded also-ran, eclipsed by DeepSeek and worth barely more than $4 billion. This week, the Beijing startup behind the Kimi chatbot is in talks to raise fresh capital at a valuation of as much as $50 billion — a roughly twelvefold jump — ahead of a planned Hong Kong initial public offering, according to a Bloomberg report confirmed by multiple outlets.
The proposed round, which Moonshot expects to begin discussing with investors in August, would be its last before going public. It comes on the back of one release: Kimi K3, a 2.8-trillion-parameter open-weight model that landed on July 16 and promptly rattled both Silicon Valley and Moonshot's Chinese rivals. Independent leaderboards from Artificial Analysis, Vals AI and others have ranked K3 among the top three models in the world — beaten only, in some indexes, by Anthropic's Claude Fable and OpenAI's GPT-5.6 — while undercutting both on price.
The financial turnaround has been just as steep. Moonshot's annual recurring revenue reached $300 million in June, up from roughly $100 million in March, with daily revenue reportedly surging sixfold after K3 shipped. The $50 billion target would follow a current financing round expected to value the company at about $31.5 billion before new money — itself a leap from around $20 billion in May and roughly $4.3 billion at the end of 2025.
The comeback, and the crush
The appetite is real enough to have broken things. Just two days after K3's launch, Moonshot paused paid subscriptions for new consumer users after demand outstripped its available compute, though its API stayed live. It was an awkward flex — the kind of capacity failure that signals a hit product and a company running short of GPUs at the same time.
Moonshot, founded in 2023 by 33-year-old Carnegie Mellon graduate Yang Zhilin, a former Meta and Google researcher, has assembled a backer list that reads like a who's-who of Chinese tech: Alibaba, Tencent, Meituan, IDG Capital, China Mobile and the Beijing AI Industry Investment Fund. To clear the path to a domestic-friendly listing, the company is racing to dismantle its offshore "red-chip" corporate structure by the end of July and has already circulated a shareholder resolution seeking approval for a Hong Kong IPO within six months.
The reason the model matters so much to the valuation is that it reframes what Chinese labs are capable of. "Moonshot AI is going toe to toe with Anthropic and OpenAI with far, far fewer resources," wrote Nathan Lambert, the Ai2 researcher whose Interconnects newsletter is closely followed by the field. "It is clearly the strongest open model ever released." Lambert argued that K3 has narrowed the U.S.-to-China performance gap from a debated six-to-nine months to something closer to three-to-five.
A capital-markets moment for Chinese AI
Moonshot is not making its move in a vacuum. Hong Kong has become the hottest venue on earth for AI listings: rivals Zhipu and MiniMax went public on consecutive days in January, with MiniMax doubling on its debut and raising $620 million, and the exchange topped global IPO league tables in the first quarter with roughly $14 billion raised across 40 deals. A Moonshot listing at anything near $50 billion would be the marquee name of the wave.
It also crystallizes a strategic bet that reaches all the way to the top of the Chinese state. At the World AI Conference the same week K3 shipped, Xi Jinping delivered a keynote committing China's AI ecosystem to open-source development and global diffusion — a rare high-level endorsement of exactly the strategy Moonshot is monetizing. The logic is that giving models away builds distribution first and profit later, the same playbook China ran in solar, EVs and manufacturing.
Not everyone sees open weights as an unalloyed good. "Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called 'accelerationists' so excited about open-weight models," wrote Dean Ball, a policy researcher now at OpenAI, arguing that free frontier models compress the margins — and terminal valuations — of the labs that spend billions to build them.
Why it matters
A $50 billion price tag on a three-year-old lab that was nearly written off in early 2026 says as much about China's capital markets as about Moonshot. With U.S. exchanges effectively closed to Chinese AI names and Beijing signaling policy support, Hong Kong is becoming the default listing venue for a national industry — and public-market money is now chasing open-weight labs that give their core product away. Moonshot is the test case for whether that model can support a mega-valuation, or whether Ball's warning about thin margins eventually bites. Either way, the release of K3's free weights on July 27 will hand every competitor on earth a frontier-class model, accelerating diffusion that no export control can easily stop.
What to watch
Watch whether the July 27 weight release actually happens on schedule, and how Alibaba's promised 2.4-trillion-parameter Qwen 3.8 and DeepSeek's V4 reshuffle the leaderboard. Watch the August funding talks for named lead investors and whether the $50 billion pre-money holds. And watch Hong Kong's regulators: a listing "within six months" implies a filing by early 2027, with Moonshot's revenue trajectory — and its compute crunch — under fresh scrutiny.
“Moonshot AI is going toe to toe with Anthropic and OpenAI with far, far fewer resources. It is clearly the strongest open model ever released.”— Nathan Lambert, Researcher, Ai2